Comparisons between PBL and non-PBL medical schools on problem-solving ability often show no differences. This could be either due to the fact that no difference in problem-solving skills exists or that the instruments used are inadequate. In this study a key-feature approach case-based examination was used to compare two medical schools in the Netherlands, one of which has a PBL curriculum (Maastricht) and one which has a program half way a transition from a non-PBL towards a PBL curriculum (Groningen). Differences were found both in proficiency scores and in the pattern of response times, both supporting the assumption that a PBL approach would lead to a higher level of problem solving ability. The effect size, however, is not as large as originally assumed by the PBL proponents. Conclusions must be drawn with caution, but it seems likely that a test based on large numbers of short cases is the most sensitive in detecting differences in problem solving ability between students of different curricula.